[mlir][memref][spirv] Add conversion for memref.extract_aligned_point… #86750
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
@llvm/pr-subscribers-mlir-spirv @llvm/pr-subscribers-mlir Author: Md Abdullah Shahneous Bari (mshahneo) Changes…er_as_index to SPIR-V Converts memref.extract_aligned_pointer_as_index to spirv.ConvertPtrToU. Index conversion is done based on 'use-64bit-index' option. Full diff: https://github.com/llvm/llvm-project/pull/86750.diff 2 Files Affected:
diff --git a/mlir/lib/Conversion/MemRefToSPIRV/MemRefToSPIRV.cpp b/mlir/lib/Conversion/MemRefToSPIRV/MemRefToSPIRV.cpp
index 81b9f55cac80f7..0ec3ad700fe807 100644
--- a/mlir/lib/Conversion/MemRefToSPIRV/MemRefToSPIRV.cpp
+++ b/mlir/lib/Conversion/MemRefToSPIRV/MemRefToSPIRV.cpp
@@ -308,6 +308,17 @@ class CastPattern final : public OpConversionPattern<memref::CastOp> {
}
};
+/// Converts memref.extract_aligned_pointer_as_index to spirv.ConvertPtrToU.
+class ExtractAlignedPointerAsIndexOpPattern
+ : public OpConversionPattern<memref::ExtractAlignedPointerAsIndexOp> {
+public:
+ using OpConversionPattern::OpConversionPattern;
+
+ LogicalResult
+ matchAndRewrite(memref::ExtractAlignedPointerAsIndexOp extractOp,
+ OpAdaptor adaptor,
+ ConversionPatternRewriter &rewriter) const override;
+};
} // namespace
//===----------------------------------------------------------------------===//
@@ -922,6 +933,20 @@ LogicalResult ReinterpretCastPattern::matchAndRewrite(
return success();
}
+//===----------------------------------------------------------------------===//
+// ExtractAlignedPointerAsIndexOp
+//===----------------------------------------------------------------------===//
+
+LogicalResult ExtractAlignedPointerAsIndexOpPattern::matchAndRewrite(
+ memref::ExtractAlignedPointerAsIndexOp extractOp, OpAdaptor adaptor,
+ ConversionPatternRewriter &rewriter) const {
+ auto &typeConverter = *getTypeConverter<SPIRVTypeConverter>();
+ Type indexType = typeConverter.getIndexType();
+ rewriter.replaceOpWithNewOp<spirv::ConvertPtrToUOp>(extractOp, indexType,
+ adaptor.getSource());
+ return success();
+}
+
//===----------------------------------------------------------------------===//
// Pattern population
//===----------------------------------------------------------------------===//
@@ -929,10 +954,11 @@ LogicalResult ReinterpretCastPattern::matchAndRewrite(
namespace mlir {
void populateMemRefToSPIRVPatterns(SPIRVTypeConverter &typeConverter,
RewritePatternSet &patterns) {
- patterns.add<AllocaOpPattern, AllocOpPattern, AtomicRMWOpPattern,
- DeallocOpPattern, IntLoadOpPattern, IntStoreOpPattern,
- LoadOpPattern, MemorySpaceCastOpPattern, StoreOpPattern,
- ReinterpretCastPattern, CastPattern>(typeConverter,
- patterns.getContext());
+ patterns
+ .add<AllocaOpPattern, AllocOpPattern, AtomicRMWOpPattern,
+ DeallocOpPattern, IntLoadOpPattern, IntStoreOpPattern, LoadOpPattern,
+ MemorySpaceCastOpPattern, StoreOpPattern, ReinterpretCastPattern,
+ CastPattern, ExtractAlignedPointerAsIndexOpPattern>(
+ typeConverter, patterns.getContext());
}
} // namespace mlir
diff --git a/mlir/test/Conversion/MemRefToSPIRV/memref-to-spirv.mlir b/mlir/test/Conversion/MemRefToSPIRV/memref-to-spirv.mlir
index 10c03a270005f1..bc2af8b6edadcc 100644
--- a/mlir/test/Conversion/MemRefToSPIRV/memref-to-spirv.mlir
+++ b/mlir/test/Conversion/MemRefToSPIRV/memref-to-spirv.mlir
@@ -1,4 +1,5 @@
-// RUN: mlir-opt --split-input-file --convert-memref-to-spirv="bool-num-bits=8" --cse %s | FileCheck %s
+// RUN: mlir-opt --split-input-file -pass-pipeline="builtin.module(convert-memref-to-spirv{bool-num-bits=8}, cse)" %s | FileCheck %s
+// RUN: mlir-opt --split-input-file -pass-pipeline="builtin.module(convert-memref-to-spirv{bool-num-bits=8 use-64bit-index=true}, cse)" %s | FileCheck --check-prefix=CHECK64 %s
// Check that with proper compute and storage extensions, we don't need to
// perform special tricks.
@@ -414,6 +415,43 @@ func.func @cast_to_static_zero_elems(%arg: memref<?xf32, #spirv.storage_class<Cr
}
+// -----
+
+module attributes {
+ spirv.target_env = #spirv.target_env<#spirv.vce<v1.0, [Kernel, Int64, Addresses], []>, #spirv.resource_limits<>>
+} {
+// CHECK-LABEL: func @extract_aligned_pointer_as_index_kernel
+func.func @extract_aligned_pointer_as_index_kernel(%m: memref<?xf32, #spirv.storage_class<CrossWorkgroup>>) -> index {
+ %0 = memref.extract_aligned_pointer_as_index %m: memref<?xf32, #spirv.storage_class<CrossWorkgroup>> -> index
+ // CHECK: %[[I32:.*]] = spirv.ConvertPtrToU {{%.*}} : !spirv.ptr<f32, CrossWorkgroup> to i32
+ // CHECK64: %[[I64:.*]] = spirv.ConvertPtrToU {{%.*}} : !spirv.ptr<f32, CrossWorkgroup> to i64
+ // CHECK: %[[R:.*]] = builtin.unrealized_conversion_cast %[[I32]] : i32 to index
+ // CHECK64: %[[R:.*]] = builtin.unrealized_conversion_cast %[[I64]] : i64 to index
+
+ // CHECK: return %[[R:.*]] : index
+ return %0: index
+}
+}
+
+// -----
+
+module attributes {
+ spirv.target_env = #spirv.target_env<#spirv.vce<v1.0, [Shader, Int64, Addresses], []>, #spirv.resource_limits<>>
+} {
+// CHECK-LABEL: func @extract_aligned_pointer_as_index_shader
+func.func @extract_aligned_pointer_as_index_shader(%m: memref<?xf32, #spirv.storage_class<CrossWorkgroup>>) -> index {
+ %0 = memref.extract_aligned_pointer_as_index %m: memref<?xf32, #spirv.storage_class<CrossWorkgroup>> -> index
+ // CHECK: %[[I32:.*]] = spirv.ConvertPtrToU {{%.*}} : !spirv.ptr<!spirv.struct<(!spirv.rtarray<f32>)>, CrossWorkgroup> to i32
+ // CHECK64: %[[I64:.*]] = spirv.ConvertPtrToU {{%.*}} : !spirv.ptr<!spirv.struct<(!spirv.rtarray<f32>)>, CrossWorkgroup> to i64
+ // CHECK: %[[R:.*]] = builtin.unrealized_conversion_cast %[[I32]] : i32 to index
+ // CHECK64: %[[R:.*]] = builtin.unrealized_conversion_cast %[[I64]] : i64 to index
+
+ // CHECK: return %[[R:.*]] : index
+ return %0: index
+}
+}
+
+
// -----
// Check nontemporal attribute
|
|
Just a friendly ping, @kuhar, @antiagainst :) |
kuhar
reviewed
Apr 15, 2024
| @@ -308,6 +308,17 @@ class CastPattern final : public OpConversionPattern<memref::CastOp> { | ||
| } | ||
| }; | ||
|
|
||
| /// Converts memref.extract_aligned_pointer_as_index to spirv.ConvertPtrToU. | ||
| class ExtractAlignedPointerAsIndexOpPattern |
There was a problem hiding this comment.
Suggested change
| class ExtractAlignedPointerAsIndexOpPattern | |
| class ExtractAlignedPointerAsIndexOpPattern final |
Comment on lines
426
to
436
| // CHECK: %[[I32:.*]] = spirv.ConvertPtrToU {{%.*}} : !spirv.ptr<f32, CrossWorkgroup> to i32 | ||
| // CHECK64: %[[I64:.*]] = spirv.ConvertPtrToU {{%.*}} : !spirv.ptr<f32, CrossWorkgroup> to i64 | ||
| // CHECK: %[[R:.*]] = builtin.unrealized_conversion_cast %[[I32]] : i32 to index | ||
| // CHECK64: %[[R:.*]] = builtin.unrealized_conversion_cast %[[I64]] : i64 to index |
There was a problem hiding this comment.
Could you group CHECK and CHECK64 together? It's a bit hard to read when interleaved
Comment on lines
444
to
454
| // CHECK: %[[I32:.*]] = spirv.ConvertPtrToU {{%.*}} : !spirv.ptr<!spirv.struct<(!spirv.rtarray<f32>)>, CrossWorkgroup> to i32 | ||
| // CHECK64: %[[I64:.*]] = spirv.ConvertPtrToU {{%.*}} : !spirv.ptr<!spirv.struct<(!spirv.rtarray<f32>)>, CrossWorkgroup> to i64 | ||
| // CHECK: %[[R:.*]] = builtin.unrealized_conversion_cast %[[I32]] : i32 to index | ||
| // CHECK64: %[[R:.*]] = builtin.unrealized_conversion_cast %[[I64]] : i64 to index |
| // ----- | ||
|
|
||
| module attributes { | ||
| spirv.target_env = #spirv.target_env<#spirv.vce<v1.0, [Kernel, Int64, Addresses], []>, #spirv.resource_limits<>> |
There was a problem hiding this comment.
ConvertPtrToU requires the PhysicalStorageBufferAddresses capability, no?
| @@ -1,4 +1,5 @@ | ||
| // RUN: mlir-opt --split-input-file --convert-memref-to-spirv="bool-num-bits=8" --cse %s | FileCheck %s | ||
| // RUN: mlir-opt --split-input-file -pass-pipeline="builtin.module(convert-memref-to-spirv{bool-num-bits=8}, cse)" %s | FileCheck %s | ||
| // RUN: mlir-opt --split-input-file -pass-pipeline="builtin.module(convert-memref-to-spirv{bool-num-bits=8 use-64bit-index=true}, cse)" %s | FileCheck --check-prefix=CHECK64 %s |
There was a problem hiding this comment.
Suggested change
| // RUN: mlir-opt --split-input-file -pass-pipeline="builtin.module(convert-memref-to-spirv{bool-num-bits=8 use-64bit-index=true}, cse)" %s | FileCheck --check-prefix=CHECK64 %s | |
| // RUN: mlir-opt --split-input-file -pass-pipeline="builtin.module(convert-memref-to-spirv{bool-num-bits=8 use-64bit-index=true}, cse)" %s \ | |
| // RUN: | FileCheck --check-prefix=CHECK64 %s |
mshahneo
added a commit
to mshahneo/mlir-extensions
that referenced
this pull request
Apr 18, 2024
…ew SPIR-V pipeline. The patch has a upstream PR pending review: llvm/llvm-project#86750. The patch can be removed once the PR gets merged and LLVM version is updated. The test cases in this PR only provides lowering from func to spirv. XeGPU test cases using the full pipeline is coming as part of next PR from Dimple (drprajap).
Merged
6 tasks
mshahneo
added a commit
to mshahneo/mlir-extensions
that referenced
this pull request
Apr 18, 2024
…ew SPIR-V pipeline. The patch has a upstream PR pending review: llvm/llvm-project#86750. The patch can be removed once the PR gets merged and LLVM version is updated. The test cases in this PR only provides lowering from func to spirv. XeGPU test cases using the full pipeline is coming as part of next PR from Dimple (drprajap).
mshahneo
added a commit
to mshahneo/mlir-extensions
that referenced
this pull request
Apr 18, 2024
…ew SPIR-V pipeline. The patch has a upstream PR pending review: llvm/llvm-project#86750. The patch can be removed once the PR gets merged and LLVM version is updated. The test cases in this PR only provides lowering from func to spirv. XeGPU test cases using the full pipeline is coming as part of next PR from Dimple (drprajap).
…er_as_index to SPIR-V Converts memref.extract_aligned_pointer_as_index to spirv.ConvertPtrToU. Index conversion is done based on 'use-64bit-index' option.
ca371a8 to
f120ca2
Compare
|
I am so sorry, Jakub (@kuhar). For some reason, I completely forgot about this. Updated the PR to address your comments. |
|
Hi @kuhar , just a friendly ping :) |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
…er_as_index to SPIR-V
Converts memref.extract_aligned_pointer_as_index to spirv.ConvertPtrToU. Index conversion is done based on 'use-64bit-index' option.